• Aucun résultat trouvé

Quantum Interference and the Spin Orbit Interaction in Mesoscopic Normal-Superconducting Junctions

N/A
N/A
Protected

Academic year: 2021

Partager "Quantum Interference and the Spin Orbit Interaction in Mesoscopic Normal-Superconducting Junctions"

Copied!
28
0
0

Texte intégral

(1)

HAL Id: jpa-00247201

https://hal.archives-ouvertes.fr/jpa-00247201

Submitted on 1 Jan 1996

HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

Quantum Interference and the Spin Orbit Interaction in Mesoscopic Normal-Superconducting Junctions

Keith Slevin, Jean-Louis Pichard, Pier Mello

To cite this version:

Keith Slevin, Jean-Louis Pichard, Pier Mello. Quantum Interference and the Spin Orbit Interaction

in Mesoscopic Normal-Superconducting Junctions. Journal de Physique I, EDP Sciences, 1996, 6 (4),

pp.529-555. �10.1051/jp1:1996228�. �jpa-00247201�

(2)

Quantum Interference and the Spin Orbit Interaction in

Mesoscopic Normal-Superconducting Junctions

Keith Slevin

(~),

Jean-Louis Pichard (~>*) and Pier A.

Mello (~)

(~) Service de

Physique

de

l'ktat Condens4, CEA-Saclay,

91191 Gif-sur-Yvette, France

(~) Instituto de Fisica, UNAM,

Apartado

Postal 20-364, 01000 Mexico D-F-

(Received

18

July

1995, revised 14 November 1995, accepted 4

January 1996)

PACS.74.80.Fp

Points contacts; SN and SNS

junctions

PACS.74.50.+r

Proximity effects,

weak

links,

tunneling phenomena, and

Josephson

effects

PACS.72.10.Bg

General formulation of transport

theory

Abstract, We calculate the quantum correction to the classical conductance of a disordered

mesoscopic normal-superconducting (NS) junction

in which the electron

spatial

and spin

degrees

of freedom are

coupled by

an

appreciable

spin orbit interaction. ~fe use random matrix

theory

to describe the scattering in the normal part of the junction and consider both

quasi-ballistic

and diffusive

junctions.

The

dependence

of the

junction

conductance on the

Schottky

barrier

transparency at the NS interface is also considered. We find that the quantum correction is

sensitive to the

breaking

of spin rotation symmetry even when the junction is in a magnetic

field and time reversal symmetry is broken. We demonstrate that this

sensitivity

is due to

quantum interference between

scattering

processes which involve electrons and holes

traversing

closed

loops

in the same direction. We explain

why

such processes are sensitive to the spin

orbit interaction but not to a

magnetic

field.

Finally

we consider the effect of the spin orbit interaction on the

phenomenon

of "reflectionless

tunnelhng."

1, Introduction

In mesoscopic

samples

of disordered normal metals at low

temperatures

it is

possible

to observe

a quantum correction to the classical Boltzmann conductance

ill.

Here

mesoscopic

means that within the

sample,

the interaction of a

single

electron with other

degrees

of

freedom,

such as other

electrons, phonons, magnetic impurities

etc., can be

neglected.

Such a situation can be realised in semiconductor and metal nanostructures cooled to milli-Kelvin

temperatures.

The

origin

of the

quantum

correction is

quantum

interference between time reversed

scattering paths. (By

"time reversed

path"

we mean that the electron follows the same

trajectory

but

m the

opposite sense.)

The correction is sensitive to the

breaking

of spin rotation

symmetry

and is

suppressed by

the

breaking

of time reversal

symmetry

[2]. The

spin

orbit interaction has the

important property

of

breaking spin

rotation

symmetry

while

preserving

time reversal

symmetry.

As an

example,

the two

probe

conductance G

=

(e~/h)g

of a

quasi-one

dimensional meso-

scopic wire can be

expressed

in the form g

=

gel

+

bg

[3]. The classical conductance of the wire

is

O(N):

gC~

=

N/(1+ s)

where N is the number of

scattering

channels and s

=

Lli

is the

(*)

Author for

correspondence (e-mail: pichard©amoco.saclay.cea.fr)

©

Les

#ditions

de

Physique

1996

(3)

ratio of the

length

L of the wire to the elastic mean free

path

I. In the absence of an

applied magnetic

field the

quantum

correction

bg,

also called the "weak localisation

correction",

is

O(N°).

More

exactly,

in the diffusive

regime

1 < s < N we have

bg

=

-2/3

if the

spin

orbit interaction is

negligible

and

bg

=

+1/3

if it is not. In a

magnetic

field sufficient to break time reversal

symmetry

the

quantum

correction is

suppressed

and becomes

O(1IN).

More

recently,

a quantum correction to the classical conductance of a disordered mesoscopic normal

superconducting (NS) junction

was observed [4]. The correction is most

pronounced

in

junctions

where the transparency T of the

Schottky

barrier at the NS interface is low T < 1 and the

length

of the normal part is such that Ts m 1. Under these conditions a

phenomenon

known as "reflectionless

tunnelling"

occurs. In [5] it is

suggested

that

quantum

interference is

responsible

for this effect. The

type

of

scattering paths

involved in this

quantum

interference contain a

segment

where the

path

of an electron

(hole)

incident on the NS

boundary

is

subsequently

retraced

by

an Andreev reflected hole

(electron.)

The enhancement of the

junction

conductance can be as much as several times

greater

than the classical conductance.

In this paper we

investigate

the effect of the

coupling

of the spin and

spatial degrees

of

freedom of the electrons

by

an

appreciable spin

orbit interaction has on this

quantum

correction.

As

previously mentioned,

this has an

important

effect on the

quantum

correction in a normal metal. As the electron diffuses

through

the

metal,

the

spin

orbit

coupling

causes a simultaneous diffusion of the direction of its

spin.

Interference between different

spin

states is now

possible

and this modifies the interference

responsible

for the weak localisation correction. What role

then,

if any, does this

spin

diffusion

play

in the NS

junction?

At first

sight

the answer to this

question

would be appear to be none at all. The

argument

is as follows: the

spin

rotations of an electron and an Andreev reflected

hole,

which traverse time reversed

paths,

are

exactly opposite

and cancel each other. Since interference between

scattering

processes

involving

such

paths

is believed to be

responsible

for the quantum cor-

rection,

it should be insensitive to the spin orbit interaction. As we shall demonstrate below this is too

simple

and in fact the

spin

orbit interaction does affect the

quantum

correction to the classical conductance of an NS

junction.

We have found the flaw in the

argument

is that

it is not

only

processes

involving

electrons and holes

traversing

time reversed

paths

which are

responsible

for the quantum correction. Processes in which electrons and holes traverse

loops

m the same sense, what we shall call here identical

paths,

also contribute and in this case the

spin

rotations do not cancel each other. The most

striking

consequence of the existence of interference

involving

such

paths

is that the quantum correction is sensitive to the

spin

orbit interaction even in a

magnetic

field.

In the normal metal we have seen that the

quantum

correction to the classical conductance

is

suppressed

when time reversal symmetry is broken in an

applied magnetic

field. This is

easily

understood since the symmetry between time reversed

paths

is broken when the time reversal

symmetry

is broken. In the NS

junction however,

the contribution of identical

paths

is still present even m a

magnetic field;

the electron and hole carry

opposite charges,

and

so the Aharonov-Bohm

phases

that

they

accumulate as

they

follo1&~ such

trajectories

cancel.

Moreover,

since the electron and hole

undergo

identical spin

rotations,

as

they

follow identical

paths,

this contribution is sensitive to the

spin

diffusion induced

by

the

spin

orbit interaction.

The paper is

organised

as follows. In Sections 2-4 1&~e deal with some necessary

preliminaries:

the

Bogolubov

de Gennes

equations, scattering theory

and conductance formulae for the NS

junction, generalising

the standard

theory

to take into account the spin orbit interaction.

In Section 5 1&~e

investigate

a

quasi-ballistic

NS

junction,

that is to say a

junction

whose

length

is shorter than the mean free

path

but

long enough

so that its

scattering

matrix is well described

by

a certain random

phase approximation

[6]. We express the

junction

conductance

(4)

Table I. The various ensembles

for

which

GNS

is calculated. The abbreviation TRS means time reversal

symmetry

and SRS

spin

rotation

symmetry.

Ensemble TRS SRS

Orthogonal

yes yes

Unitary

I broken yes

Unitary

II broken broken

Symplectic

yes broken

Table II. The

quantum

correction

bGNs

"

(2e~/h)bgNs for

the

quasi-ballistic

NS

junction

for

the ensembles listed in Table I.

Ensemble

bgNs

T

= 1 T < I

Orthogonal/Symplectic -2~Tsf(r)

-4Ns

+Nsr~

Unitary

-2s

f(r)

-4s

+sr~

Unitary

II +s

f(r)

+2s

-sr~ /2

GNS

as a sum of t~vo terms

GNS

"

~)

gNs "

~) (gis

+

~~NS)

(~~

a classical conductance

g(~

and a

quantum

correction

bgNs

due to

quantum

interference. We determine the

quantum

correction for the four ensembles listed in Table I and the results are

presented

in Table II. In Section 6 we present a semiclassical

interpretation

of these i«esults which

permits

us to

identify

the

type

of

scattering paths

which interfere to

produce

the

quantum

correction. In

agreement

with reference

[5],

we find that there is an

important

contribution due to interference between processes

involving

electrons and holes

traversing

time reversed

paths,

but that there is also a second contribution from processes

involving

electrons and holes

traversing

identical

paths.

We

explain why

this contribution is sensitive to the

breaking

of spin rotation

symmetry

but not to the

breaking

of time reversal

symmetry.

After

reaching

a firm

understanding

of the

quasi-ballistic junction

we consider the diffusive

regime, looking

first at

junctions

without a

Schottky

barrier at the NS interface m Section 7.

In zero

magnetic

field the quantum correction is known to be of

O(N°) [7-9].

In a

magnetic

field the authors of reference [10] find that the correction

though

smaller is still

O(N°).

The effect of the spin orbit interaction was not considered in reference

[10j

and so we repeat their calculation

taking

it into account. We find that

breaking

spin rotation

symmetry multiplies

the correction

by

a factor of minus one

half, regardless

of whether time reversal

symmetry

is

broken or not.

In Section 8 we discuss the dramatic enhancement of the

junction conductance,

kno1&~n as

"reflectionless

tunnelling,"

which is observable when r < 1 and rs m 1. Under these conditions the contribution of time reversed

paths

is

O(N)

and dominates that of identical

paths

which

is

O(N°)

so that the reflectionless

tunnelling

effect

is,

m a first

approximation,

insensitive to the spin orbit interaction. This conclusion has been confirmed

by

carrying out a numerical

simulation of a

junction

under the relevant conditions.

(5)

2~

Bogolubov

de Gennes

Equations

and the

Spin

Orbit Interaction

The

Bogolubov

de Gennes

(BdG) equations

[11]

appropriate

for a metal where the electrons'

spatial

and

spin degrees

of freedom are

coupled by

a

significant spin-orbit interaction,

are

l~~ ~*'

j-~~ j-

~h

~/,e " f

~h

~fie

(2)

' ~

where

He

=

Ho EF

+ Q

Here

Ho

e

Ho (r,

a,

p)

is the

single

electron Hamiltonian of the metal

incorporating

the spin- orbit interaction and

EF

is the Fermi energy. In

deriving

this

equation,

it is assumed that an

attractive point like interaction exists between the electrons:

The full

interacting

Hamiltonian is reduced to an effective

non-interacting

Hamiltonian

by introducing

effective

potentials

~(r)

=

%*(r, I)*(r,1) (4)

Q(r,

a,/~)

=

lfi(1 2ba,~)*lr, a)*t(r,

/~) 15)

where the overline indicates a thermal average with respect to the Fermi distribution function and §/ is the usual field

operator

appearing in the second

quantised

formulation of the inter-

acting

electron

problem.

The time reversal

operator T,

which appears in the BdG

equations

has the form

T =

pC (6)

p =

Say

=

(~ ii)

~~

with C the

operation

of

complex conjugation.

The

eigenstates

of the BdG

equation

describe the excitations of the

interacting

electron

system.

The

meaning

of the electron ifi~ and ifi~

wavefunctions can be seen

by writing

the field operator as

*lr, a)

=

~j (ifi[lr, a)n~~ Tlifillr:

a)In~)j

18)

where ~n is the annihilation operator ofthe excitation labelled n. The

corresponding eigenvalue

en of the BdG

equations corresponds

to the energy of the excitation. With the aid of

(8)

the

occupation

of the

eigenstates

of

He

for a

given

excitation can be determined.

3.

Scattering Theory

for the NS Junction

In this section 1&~e

develop

the

scattering theory appropriate

to the normal

superconducting junction

shown

schematically

in

Figure

1. We treat the

scattering

at the NS interface within

Andreev's

approximation.

We suppose that any

impurities

in the

system

are m the region indicated

by shading

in

Figure 1,

and we make the

approximation

that the

magnetic

field is

zero

everywhere except

in this disordered region. This is reasonable for the low fields of interest which affect the interference between electrons and holes m the normal part of the

junction.

(6)

NORMAL SUPERCONDUCTOR MBTAL

RANDOMPOTBNTIAL SCHOTTKYBARRW

Fig.

1. A schematic of the NS

junction

for which the scattering

theory

is

developed

in Section 3.

3.I. SCATTERING MATRICES FOR ELECTRONS AND HOLES. To facilitate the

explanation

of the

formalism,

it is

helpful

to

develop

the

scattering theory

for a definite model. We have chosen a lattice

tight binding model,

the same model which we will use later m numerical

simulations. An exact

analogous explanation

is also

possible

for a continuum model.

We consider a cubic lattice and take into account nearest

neighbour

interactions

only.

We denote

by ~~(x,

y, z,

a)

the

amplitude

that the electron is in an s- orbital at

lx,

y,

z)

with

spin

a and

similarly

ifi~

lx,

y, z,

a)

for a hole. We include in the Hamiltonian a

spin

orbit term which arises from the Zeeman

coupling

of the electron

spin

with the effective

magnetic

field felt

by

the electron as it moves in the

spatially varying potential

of the lattice. We

ignore

the direct Zeeman

coupling

of the

spin

to any external

magnetic

field. A

simple

calculation shows that the Hamiltonian has the form:

< X>Y> Z,

a(tie(I,

Y,Z,JL > "

E0~a,v

< x, Y, z,

alHe lx i,

Y, z, J1> =

vS,~

< x, v, z,

alHe lx, i,

z, J1 > =

vl,~

~~~

< X> l/, Z,

ajHelX,

l/>Z

I,

/l > #

eXp(-S£XX) Uj,~

where

vj

~ =

l§i~,~ vii[a~]~~~

U$)v =

l§b«,v Vii(ay]a,v (10)

vj,~

=

l§i~,~ vii[az]~,~

An external

magnetic

field

B, applied

in the +y direction is modelled

by

Peierl's factors in the matrix elements between nearest

neighbours

m the z direction. If a is lattice constant a =

27rBa~ /~o

where

~o

=

hle

is the flux

quantum.

The transverse dimensions are1 < x <

L~

and 1 < y <

Lg.

The Hamiltonian

(9)

can be

regarded

as a three dimensional

generalisation

of that

proposed

in reference

[12]

as a model for a two dimensional electron gas formed at the surface of a III-V semiconductor.

The relevant energy scale of the model is determined

by (~

+

l§~.

For convenience we shall set this to

unity

with the choice

= cos 6

Vi

= sin 6

(11)

Varying

the

angle 6,

we may set an

arbitrary

ratio of normal

potential coupling

Vo to spin orbit

coupling Vi,

while

keeping

the extent in energy of the

density

of states

roughly

constant. ivith this

choice, u~,uY

and u~ are all elements of

SU(2). Using

the

homomorphism

of

SU(2)

with the three dimensional rotation group

SO(3),

we can

interpret

the u's as rotations of the spin of the electron as it moves between nearest

neighbours [13].

A

product

of nearest

neighbour

matrix elements

along

a

path

will have the form

exp(14l)u

where 4l is the Aharonov Bohm

(7)

phase picked

up

by

the

electron,

and u E

SU(2)

is the rotation of the electron's

spin

as it traverses the

path.

The Hamiltoman

He

has the form

given

in

(9) everywhere except

in the disordered

region

located in 0 < z < L.

There,

some or all of the Hamiltonian matrix elements are

supposed

random. In reference

[12]

the

diagonal

elements were assumed to be

independently

and identi-

cally distributed,

while the

parameter

6

controlling

the spin orbit interaction was held fixed. In

references

[14,15]

a random

spin

orbit interaction was also considered. For the

present

purpose

we do not need to

specify

the

precise

distribution.

First,

we consider the

scattering

of electrons incident at an energy E

=

EF

+ e. To the left

of the disordered section we

expand

the electron wavefunction ifi~ in terms of the Bloch states of

(9)

with energy E.

~~(x,

y, z,

a)

=

~j a+n~+»(x,y, a) exp(+ik~z)

+

~j a-~ifi-»(x,y, a) exp(-ik~z)

n;Zmkm<0 tZmkm=0

(12) ifi-~(x,Y,a)

=

£P~,~ifi+~lL~

x +

i,Y,J1) l13)

~

As z

~ -cc, far from the disordered

region,

we

impose

the

boundary

condition that the allowed

states consist

exclusively

of

incoming

and

outgoing propagating

waves. Thus in z < 0 states

with Zm

k~

> 0 are excluded. We denote

by

2N the number of

"open channels",

I-e- states with Zm

kn

= 0 that carry a

positive probability

current in the +z

direction;

there are an

equal

number

carrying

current in the -z direction. We label these states so that +n carries a current

m the +z direction and -~ a current in the -z direction. After a suitable normalisation of

the transverse wavefunctions

(see Appendix A)

the electric current due to electrons at the left

of the disordered section is

Ie

"

) ~j l~+~l~ l~-~l~ (14)

~,Zmkm=0

The

boundary

condition as z ~ -cc,

imposed above,

ensures that

only

open

channels,

and not "closed channels" with Zm

k~ # 0,

contribute to the current. A similar

expansion

may be made on the

right

in terms of a set of coefficients

(a[~, a[~)

with the

boundary

condition that

~ve admit

only

those states ~vith Zm

k~

> 0.

Thus,

far to the

right

of the disordered

section,

as z ~ +cc, the allowed states again consist

exclusively

of

incoming

and

outgoing propagating

waves.

The 4N x 41V

scattering

matrix for electrons

Se

relates the 4N

incoming

flux

amplitudes

at the

left,

a+

=

(a+»,Zm

km

=

0)

and the

right a[

=

(a[~,Zm

km =

0)

with the 4N

outgoing

flux

amplitudes

at the left a-

=

(a-»;Zm kn

=

0)

and the

right a[

=

(a[~,Zm

km

=

0)

Se ~/

"

~T (15)

~- ~+

The matrix

Se

has the structure

s~

l~e

te ~~

jilt)

T~

m terms of the 2N x 2N reflection and transmission matrices for left incidence

(r~,t~)

and

right

incidence

(r[, t[).

Since we are

considering

time

independent scattering,

the currents to the left and

right

of the disordered section must be

equal,

and therefore it follows that S~ is unitary. There is an

(8)

additional restriction on

Se when,

in the absence of an

applied magnetic field,

the Hamiltonian is time reversal invariant ie.

[He,

2~j = 0 with T given m

equation (6).

For a suitable choice of

transverse wavefunctions

(see Appendix A) Se

will then

satisfy

~~ ~~ -~2N ~~ ~~ ~N

~~~~

where

12N

means the 2N x 2N unit matrix. This can be written in the

equivalent

form

re =

-r) r[

=

-(r[)~ (18)

te "

+(t[)~

The

simplicity

of these

relations, compared

with for

example

those of reference

[16],

is related to the presence of p in

equation (13) (see Appendices

A and

B)

We now turn to the

scattering

matrix

Sh

for the holes. From the BdG

equations

we can see

that the hole wavefunction ifi~ evolves

according

to

Hh~~

" ih

~

ifi~

(19)

where

Hh

is given

by

Hh(+B]

=

TH~[+B]T

=

-He j-B] (20)

If ~fi~ describes the

scattering

of a hole with excitation energy +e in a field +B then

T~fi~

describes the

scattering

of an electron at energy -e also in field +B. We shall make use of this in two ways.

Firstly,

outside the disordered

region

B

= 0 and

Hh

=

-He.

Thus outside the disordered

region

it is useful to

expand

~fi~ in terms of the Bloch states of

He

at energy

EF

e. At the

left,

for

example

ifi~(x,

y, z,

a)

=

£ b-»ifi+»(x,

y,

a) exp(+ik»z)

+

~j b+»ifi-»(x,

y,

a) exp(-ik~z)

~;Zm km <0 «Zmkn=0

(21)

Note that since in this

region Hh

"

-He

the

probability

currents are reversed

by comparison

with the electron case. We therefore associate the coefficient

b+~,

the flux

amplitude

for a

positive

hole

probability

current in the +z

direction,

with the wavefunction

proportional

to

exp(-ik~z).

The holes carry an

opposite

electric

charge

to that of the electrons so that

they

carry an electric current

Ih

=

~

~j (b+~[~ [b-~[~ (22)

~

wZmkn=0

By definition,

the 4N x 4N matrix

Sh

relates incoming hole

probability

currents to

outgoing probability

currents

Sh I)/

=

)/ (23)

+

The matrix

Sh

has a structure similar to that of

equation (16);

1e.

~h

"

~

~) (2~)

h ~

(9)

Secondly by rewriting Tifi~

in terms of electron flux

amplitudes,

and

recalling

that these

amplitudes

are related

by S~,

we arrive at a relation between

Se (-e, +B)

and

Sh(+e, +B)

sh(+f,+B)

"

~f _~~~ sll~f,+B) ~j~

1)~ 125)

or

rh(+e,+B)

=

-[r~(-e,+B)]*

~ll+f>+B)

~

~l~ll~f,+B)l~

th(+e,+B) j~~~

=

+[te(-e,+B)]"

t[(+e,+B)

=

+(t[(-e,+B)]*

Again

we assume here a suitable choice of transverse wavefunctions

(see Appendix A) Following

a similar line of

argument

it is

possible

to demonstrate a

relationship

between

Se (+e, +B)

and

Se (+e, -B).

3.2. ANDREEV SCATTERING AT THE NS INTERFACE. In this section we outline the cal-

culation of the coefficients of Andreev reflection

(17]

at the NS interface. These are

essentially unchanged by

the introduction of a

spin

orbit interaction in the materials

forming

the

junction.

We assume

throughout

that

/lo

<

EF,

a condition which is realised in

practice.

Far from the NS interface m the normal metal the

superconducting

gap /l ~ 0. On the other

hand,

far from the NS interface in the

superconductor,

/l ~

/lo exp(i~)

where

/lo

is real. In

general

the reflection coefficients will

depend

on the

precise

form of /l in the transition

region

near the

junction.

For a

point

contact

junction, however,

it is

permissible

to assume a

simple step

model

=

~o exp(i~)

~~~~

We are interested in the situation where the energy E

=

EF

+ e of the incident electron is in the energy gap of the

superconductor EF

< E <

EF

+

/lo. Anticipating

somewhat in order to avoid unnecessary

algebra,

we find the electron is

mainly

reflected as a hole like excitation. A

solution of the BdG

equation

in the normal metal

corresponding

to this is

l~~

"

~XP(S~~~Z)~fi~(X>Y>~)

~

~( ~XP(S~~ ~Z)~fi~~(X>Y,~) (2~)

e

where

r)~

denotes the matrix of Andreev reflection

amplitudes

and the

superscript (+)

refer to Bloch states with energies

El+)

=

EF

+ e. The first term describes an excitation where

an electron above the Fermi level is incident from the left in channel n. The second term

corresponds

to an excitation in which an electron below the Fermi level is

annihilated,

i-e-

to a reflected "hole" with

opposite velocity

and spin to that of the incoming electron. Since

/lo

<

EF

we can to a

good approximation

ignore the difference between

k(~~

and

k(

and

similarly

the differences between the transverse wavefunctions.

Requiring

that the wavefunction and its derivative be continuous at the

boundary

of the

superconductor

leads to

r)~

=

iexp(-i~)

,

r$~

=

iexp(+i~) (29)

Here we have made the further

assumption

that e <

/lo,

the limit of interest in what

follows,

and we have also given the reflection coefficient for an incident hole.

(10)

3.3. THE SCOTTKY BARRIER AT THE MS INTERFACE. in real MS

junctions,

a mismatch between the conduction bands of the two materials which make up the

junction

results in the creation of a

Schottky

barrier at the interface. This barrier

plays

an

important

role in the

physics

of the device and so we must take it into account. We shall model the

Schottky

barrier as

a

simple planar potential

barrier. At the barrier an incident

particle

may be either transmitted without a

change

of momentum or

specularly

reflected. We

neglect

any

dependence

of the reflection and transmission

probabilities

on the momentum of the incident

particle

so that the

properties

of the barrier are described

by

a

single

parameter r E

(0,1],

its

transparency.

The transmission and reflection matrices which make up the electron

scattering

matrix

St

of the barrier are

tf

=

/f 12N

#

~r ~~

~~~~

r[~

=

-fit Ml

The

precise

form of the 2N x 2N matrix MB

depends

on the choice of transverse wavefunctions.

For

subsequent analysis

we need

only note, however,

that MB is

antisymmetric

and

unitary.

The hole

scattering

matrix

St

for the barrier is related to

St

in the usual way

by (26).

3.4. COMBINATION oF SCATTERING MATRICES. It is the purpose of this

section, having

considered above the

scattering

matrices for the various

components

of the NS

junction,

to

explain

how the

scattering

matrices may be combined to find the total

scattering

matrix. We consider first a

junction

without a

Schottky

barrier. For the normal part we can write

"~~~~~

~ =

~e ~

etc. ~~~~

0 rh

are 4!V-dimensional matrices and

C+ ~

, C+ ~ ~i

(33)

~~

'

~

4N-dimensional vectors, in the notation of Section 4. For the Andreev

part

we define the 4N x 4N

scattering

matrix

S~ by

S~c[

=

c[ (34)

By

reference to Section 3 this has the form

~ ~

r)~12N

0 ~~~~

For the combined

system,

the 4N x 4N S matrix is defined

by

Sc+

= c-

(36)

In

equation (31)

we

replace

CL in terms of

c[

as in

(34);

we then

perform

the matrix multi-

plication and,

from the two

resulting equations,

eliminate

c[.

From the

expression relating

c+

and c- we extract the 4N x 4N S matrix of

(36)

to obtain

S = r +

t'S~

~,~~

t

(37)

(11)

In the electron-hole spaces, S of

equation (37)

has the structure

S=

~~~

~~~

(38)

The rhh

re~ etc.

being

2N x 2N matrices. To determine the conductance 1&~e shall need the submatrix The From

equations (37)

and

(38)

we find

Using

the structure

(35)

of

S~

we can write The as

~~~ ~ ~~~~~

l r'SA

~~

~~ ~~

For any

nonsmgular operator

D we now use the operator

identity

(18]

1

Ii

~

(41)

D

ee

l~ee ~e~ ~~~~

ee

~~ ~~~

The "

ti~te l~ ~[~te~i~shl

~~~

l~~~

For a

junction

with a

Schottky

barrier we first consider the

composition

of the

Schottky

barrier and the

superconductor.

The

scattering

matrix for this system can be obtained from

equation (37)

where r, t and t' are taken from the model

(30)

for the barrier. With the aid of the

identity (41)

and the

following

one

(18]

l~

he

~h ~~~

Dee

~h

~Dhe

~~

~~~~

~~ ~~~

~BS

T~S ~BS

"

()

eh

~he

T~)

~~~

l

~XP(~S~) (~/(2 ~)j

12N

~fS

=

_j~[BS]

~~

~~i

" 2

~/(2 ~)j

WB ~~~~

~BS

_j~[BS]

~~

hh ee

The

scattering

matrix for the

complete system

of normal part,

Schottky

barrier and supercon- ductor can now be obtained from

equation (37),

where

r',

etc. refer to the normal metal as

before,

but

S~

is

replaced by S~~

of

(45). (Note

that in

deriving Eq. (371

the structure

(35)

of

S~

1&~as not

used.)

4. Conductance Formulae

We assume that the bias

voltage

is small in comparison to

/lole,

and that the size of the super-

conducting

part is

long enough

so that there is no

quasi-particle

current in the

superconductor.

(12)

In this case, the zero temperature dc conductance

GNS

of the

normal-superconducting junction

can be described

by

a

simplified

"Landauer" formula which has been derived in reference

(19-21]

where The is the 2N x 2N matrix of electron-hole reflection

amplitudes

for the

composite

system.

There is an

important simplification

if the Hamiltonian of the

system

is time reversal invari- ant and the bias

voltage

is small in

comparison

to the Thouless energy

Ec [22],

so that the

energy

dependence

of

Se

can be

neglected.

The conductance then has the form

[23]

where

Tn

are the

eigenvalues

of

tet).

We have verified that this result still holds when

spin

rotation

symmetry

is broken

by

the

spin

orbit interaction.

In what follows we wish to compare the

quantum

conductance of the NS

junction,

calculated from

(46),

with the classical conductance of the

junction.

This latter

quantity

is determined

using the classical rule of

combining

conductances

i/g

=

i/gi

+

1/g2 148)

This

corresponds

to the addition of flux intensities as

opposed

to flux

amplitudes.

The conduc-

tance associated with the electron

traversing

the normal

part

is

2N/s

and

similarly

for the hole

m the

traversing

the normal part in the

opposite

direction. The conductance of the barrier is

2N[r))[~,

so that the classical conductance is

~j~~j~BSj2

~~~

1 +

2s~~))

[2 ~~~~

The classical conductance is insensitive to the

breaking

of time reversal and spin rotation

symmetries.

5.

Quasi-Ballistic

Junction

In this section we shall calculate the conductance of an NS

junction

to first order m s

=

Lli,

where L is the

length

of the normal part of the

junction

and I is the elastic mean free

path.

We

neglect

terms of order

s~

and above

so that result is

strictly applicable only

m the limit that s < 1. Nevertheless we shall see that the results of the calculation shed considerable

light

on the

origin

of the

quantum

interference in the device.

The 2N x 2N reflection matrix The for the

system consisting

of the normal

metal,

barrier and

superconductor

can be obtained as discussed in Section 3.4. After

expanding

to second order m

r[, r[,

which is sufficient for an evaluation of gNs to first order in s, we have

~he ~~~~~~~ ~

~~~~i~~~ii~e

+

~~~~/~~~~~~e

+

~~~~~~~~ii~~Tii~e

(50)

+

t[r))r[r))r[rf~~t~

+

t[r))r[r))r[r))t~

+

t[r))r[rf)r[r))te

+

The conductance is found

by substituting

this into

(46)

and

performing

an average over an

ensemble

ofscattering

matrices

Se, describing

a set

ofmicroscopically

different but macroscop-

ically equivalent configurations

of

impurities

m the normal part of the

junction.

(13)

In

principle,

the distribution for

Se

should be calculated from some model distribution of Hamiltomans. We shall

not, however, attempt

to do that here. Instead we will assume that the

resulting

ensemble of

scattering

matrices

Se

is distributed

according

to the "local maximum

entropy

model"

[3, 24, 25].

If the

geometry

is

quasi-1d

and the number of channels N

sufficiently large,

then results obtained with the aid of the local maximum

entropy

model are known to be identical to those obtained from the class of

microscopic

models described

by

the nonlinear

sigma

model

[26, 27].

The distribution of

Se

m the local model

depends

on

N,

s and the

symmetry

of the Hamil-

tonian,

I.e. whether or not time reversal

symmetry

is broken and whether or not

spin

rotation

symmetry

is broken. There are

four

ensembles

(Tab. I.)

The critical

strengths

of the

magnetic

field and the spin orbit interaction

separating

the various ensembles should be similar to those associated with the weak localisation effect in normal metals. The details of the distribution of S~ for the four ensembles can be found in

Appendix

C.

It will be

helpful,

when we come to discuss the

interpretation

of the

results,

to write

(50)

in the form

~

The "

~

Pi

(51)

1=1

Each term m the series

represents

the contribution of a

particular scattering

processes to The- The classical conductance is obtained

by ignoring

interference between different processes and

summing

intensities

cc

g#s

= tr

i=1 ~j p~p) (52)

The quantum correction to this classical conductance is found

by

summing the interference between different processes

bgNs

" tr

i#j ~j p~pj (53)

After carrying out the average we find that

g~8 2~'T~~'~(1~ 2'T~~'~S

~

°(S~)) (5~)

which agrees with the

expansion

of

(49)

to the order we are

considering.

The

explicit expressions

for

bgNs

are collected

together

m Table II. The function

f

of the barrier

transparency,

which

appears in the

table,

has the

explicit

form:

fir)

-

~~~l~~ i~~~l~~

ii 155)

There are two obvious

limiting

cases: T = 1

corresponding

to a

junction

without a

Schottky

barrier and r <

corresponding

to a

junction

with a

high Schottky

barrier. The first

point

to note is that the

quantum

correction is of

O(N),

the same order as the classical

conductance,

m zero field.

Secondly

for T <

1,

the conductance increases as disorder is added to the

junction.

This is the essence of the dramatic reflectionless

tunnelling

effect which we discuss

m Section 8.

Thirdly

when time reversal

symmetry

is broken

by

the

application

of a

magnetic

field the

quantum

correction is

O(N°

and not

O(1IN)

as

might

have been

expected by analogy

with the weak localisation effect in a normal metal. The

final,

and

perhaps

the most

surprising

result,

is that in a

magnetic

field the

breaking

of

spin

rotation

symmetry by

the spin orbit interaction

multiplies

the

quantum

correction

by

a factor of minus one half even

though

the

symmetry

of the

Hamiltonian,

m the sense of random matrix

theory,

is

unchanged

and remains unitary.

(14)

xmx

NORMALMBTAL SUPERCONDUCTOR

Fig.

2. An

example

of a

path

which contributes to process pi

corresponding

to an electron

(solid linel

traversing the normal part of the

junction

whose

path

is then retraced

by

the Andreev reflected

hole

(dashed line).

6. Semiclassical

Interpretation

The

importance

of

quantum

interference between processes in which the

path

of an electron

(hole)

incident on the NS

boundary

is

subsequently

retraced

by

an Andreev reflected hole

(electron)

was first

pointed

out in [5]. In the absence of a

magnetic field,

and if the bias

voltage

is small

enough,

electrons which move

along

a

path

in one

given

sense are

phase conjugated

with holes

traversing

the time reversed

path.

In a

magnetic field,

or if the bias

voltage

is

large enough,

this

phase conjugation

is

destroyed.

However we have seen that there is a

significant quantum

correction even in a

magnetic

field. There must therefore be an additional source of

quantum

interference which is not sensitive to the

breaking

of time reversal

symmetry.

As

we shall see the relevant processes involve

paths

m which an electron

(hole)

and an Andreev reflected hole

(electron)

traverse a

loop

in the same sense. In order to remain concise, we shall refer to such processes as

containing

identical

paths.

The interference

involving

such

paths

can

only

be

destroyed by applying

a

large enough

bias

voltage.

The

physical importance

of the

bias

voltage

as a

"symmetry breaking parameter"

is discussed in

[10].

Only

some of the terms in

(53)

are found to be nonzero after averaging, so that in fact

bgNs

= tr

(pip)

+

Pip(

+

psp)

+

p7p()

+

Ols~) 156)

where pj means the

jth

term in

(50.)

Consider first the interference between process pi and p5. An

example

of a

scattering path

which contributes to process pi

Pi "

t[~))te 157)

is illustrated in

Figure

2. Since we are

working

to first order in s it is sufficient to consider the motion of the electron and

holes, traversing

from one side of the normal

part

of the

junction

to the

other,

as

ballistic,

so these

trajectories

appear as

straight

lines m

Figure

2.

Examples

of

paths

which contribute to p5

P5 "

t~Tf~T~Tf)~~~iite (~~)

are illustrated in

Figures

3 and 4. To first order m s the interference between processes pi

(15)

X=Impufity

x«xb

x=xa

NORMALMBTAL SUPERCONDUCTOR

Fig.

3. An

example

of a "time reversed

path"

which contributes to process p5. The

path

of the

electron moving from xa to xb is retraced by the Andreev reflected hole as it moves from xb to xa.

Interference between this

path

and that illustrated in

Figure

2 is insensitive to the spin orbit interaction and is

suppressed

in a

magnetic

field.

X=ImpuTity

X"Xa"Xb

NORMALMBTAL SUPERCONDUCTOR

Fig.

4. An

example

of an "identical

path"

which contributes to the process p5. The electron moves around a

loop

and returns to ~a. The Andreev reflected hole traverses the

loop

m the same direction.

Interference between this

path

and that illustrated in

Figure

2 is sensitive to the

spin

orbit interaction but not to a

magnetic

field.

and p5 is

tr

(pip)

+

p5P))

"

2[r))[~[rf~~[~tr(r[r[) (59)

The

remaining

terms

involving

pi and p7 contribute

tr

(piP~

+

p7P))

"

-2[r))[~tr (r[r() (60)

In the interest of

brevity

we will concentrate on the interference between pi and p5. A very similar

analysis

is

possible

for the interference between processes pi and p7

leading

to identical

conclusions. We

proceed by relating

the

product

of electron and hole reflection matrices m

(59)

to a

product

of an electron and a hole Greens function. To

simplify

the

algebra

we shall

suppose that both the spin orbit interaction and the

magnetic

field are zero

everywhere

except

(16)

m the disordered

region.

We shall also suppose that

Ly

= 1 and

impose periodic boundary

conditions in the x direction. The Bloch states at energy E have the form

1fi2m(x,z,a)

=

exp(ikj~x)exp(ik2mz)b~~~~~~

a2m "

I

~fi2m+1(X,

Z,

a)

"

eXp(Sk(m+lX) eXp(Sk2m+1Z)b~,~2»1+1

~2m+1 "

I

(61) ifi-2m(x,z,a)

=

-exp(-ik(~x)exp(-ik2mz)b~,~_~~

a-2m =

~fi-(2m+1)(X,Z,a)

"

eXp(-Sk(m+lX)~XP(~S~2m+lZ)~a,a-(2m+1)

~-(2m+1)

i

~~~~~

k(~

=

k(m+1

"

)~°

"

~''' '~~

~~~~

and the energy and the momenta are related

by

E = 2 cos

kc

+ 2 cos

km (63

The reflection matrices for electrons and holes can be related to the

corresponding

Green's

functions as indicated in

Appendix

A.

lr~)m~

=

-i~exp(+i(km+k~)L)

~j ~fi~m(Xba)G$ (Xb,

L> ~l Xa L>

~')~fi-~ (Xa~') (64)

xaiba'

[r[]~m

=

-i~exp(-I(km+k»)L)

~j ifi[~(xba)G)(xb,L,a;xa,L,a')ifi+m(xaa') (65)

~~~~~aJ

Note

that,

for

convenience,

the states

(61)

have not been normalised to carry identical currents, proper account of this has been taken in the expressions

(64)

and

(65)

for the reflection matrices.

The

quantum

correction involves a trace over the

product

of r~ and rh.

Using

the relation with the Green's function this can be

separated

into two elements: an

integration

over the cross section

involving

the transverse wavefunctions and an average of the

product

of an electron and a hole Greens function. We will consider the second element first. This involves

evaluating

A~_~,~

~m

(Xa, Xb)

"

(~i~(Xa> L,

a-n, Xb,

L, a+m)G$(Xb, L,

a+m, Xa,

L, a-n)) (66)

Within the semiclassical

approximation,

as is

explained

in reference

[28], Chapter

12 and

13,

we can express the Greens functions as summations over

paths:

G~(Xb, L,

a+mi Xa,

L, a-n)

"

£~=~a-~b ~J ~XP(S~J

~

S~J)iUJj?+»i

?-n

(fi?)

G~

~~b>

L,

a-n, Xa,

L, a+m)

"

£j:~a-xb ~J ~XP(~S~J S~J (UJj?-n,?+»1

Substituting (67)

into

(66)

and

taking

the disorder average we find that

only

two contributions

remain:

6.I. TIME REVERSED PATHS. The

path

of the electron

moving

between xa and xb is

retraced

by

the Andreev reflected hole as illustrated in

Figure

3. The electron and hole

charges

are

opposite

but

they

traverse the

path

in

opposite

directions so the Aharonov-Bohm

phase

factors

they

accumulate do not cancel each other out. Thus this term is

important only

when the

magnetic

field is

negligible

and the Hamiltonian is time reversal

symmetric.

Références

Documents relatifs

Significance and Impact of the Study: Our results provided the first evidence on the impact of the pollutant 2,4-DNT on the indigenous bacterial commu- nity in marine sediment,

At the alpine treeline, we experimentally warmed undisturbed soils by 4 K for one growing season with heating cables at the soil surface and measured the response of net C uptake

La maîtrise des risques infectieux associés aux soins (IAS) dont les problèmes d’accidents exposant au sang (AES) le personnel de santé ainsi que les défections

Dans ce contexte, l’apport envisagé avec ce travail de thèse est de collaborer à la conception optimale d’un système de conversion de l’énergie

nonproper spinflips in POL2, correlations between the polarization and the energy of the scattered beam with its position and angle at the POL2 targets and due

Furthermore, our results indicate that number of dark trades within a predetermined time- interval provides more information regarding the (one-step-ahead) point and density forecast

تارﻴﺜﺄﺘﻝا ﻩذﻬﺒ لﺌاوﻷا نﻴﻴﻨﻌﻤﻝا نﻴﺒ نﻤ تﺎﻌﻤﺎﺠﻝاو دﻫﺎﻌﻤﻝا نﺎﻓ ﺔﻤﺎﻬﻝا نﻴدﺎﻴﻤ نﻴﺒ نﻤ ﻲﻤﻠﻌﻝا ثﺤﺒﻝا عﺎطﻗ لﻌﻝو فﻠﺘﺨﻤ ﻲﻓ لﺎﺼﺘﻻاو مﻼﻋﻹا ﺎﻴﺠوﻝوﻨﻜﺘ ﺞﺌﺎﺘﻨ قﻴﺒطﺘ

Distribution of pre-Aksumite and late pre-Aksumite sites around Wakarida and the location of the three excavated sites (site no. 44: Mangagebit) (French-Ethiopian